Sign in with
Sign up | Sign in
Your question
Solved

Graphics and System Crashes - Updated

Last response: in Systems
Share
June 15, 2011 6:11:31 PM

Problem: Most games (Portal 2, Assassin's Creed: Brotherhood, Fallout: New Vegas) crash within 5 minutes of startup. Usually they will crash the video driver at least one, the machine will recover and then hard crash (requiring reset) shortly thereafter).

Troubleshooting: I have done sustained GPU testing with Furmark, no errors or crashes. Testing with Prime95 and MemTest+ (separately) overnight, no errors or crashes. Driver cleaner used to wipe graphics drivers and then reinstalled. Finally, wiped the OS on Sunday, formatted the drive and did a clean reinstall, fully patched windows, added drivers and installed AC:B. Worked with one crash for several hours, following day, back to old problems. Similar results with Portal 2.

Rig:

Build Date: 12/27/2009

ASUS P6T Deluxe
Intel i7 920 (45nm)
MSI 5850 HD Graphics Card
6GB (3x2GB OCZ DDR3 1600)
850W Seasonic PSU
Auzentech Home Theatre HD
300GB Velociraptor
Windows 7 64 Pro

The graphics card and CPU are watercooled. Temperatures are good (peak 100F at sustained full load). All gear is in a HAF case. Monitored voltages appear to be steady on 3/5/12V rails under a variety of load conditions.

I'm at the end of any fixes I know short of throwing the contents of the case out the window and starting over. There is still plenty of life left in this system, so I'd prefer to not do that.

Any troubleshooting advice, ideas would be greatly appreciated. Thanks.

Update: On a whim, tried AC:B again 2x : Crash within 20 seconds of entering game. Pulled the Auzen card. System appears slightly more stable (dies in 5 minutes vice 5 seconds) but still experiences the hard crash.
a c 113 B Homebuilt system
June 17, 2011 3:41:25 AM

cdrikari said:
All gear is in a HAF case. Monitored voltages appear to be steady on 3/5/12V rails under a variety of load conditions.

Update: On a whim, tried AC:B again 2x : Crash within 20 seconds of entering game. Pulled the Auzen card. System appears slightly more stable (dies in 5 minutes vice 5 seconds) but still experiences the hard crash.


How did you monitor voltages? Onboard sensors are useless. unless you have a good DMM you won't know. But of course a Seasonic PSU is not too suspect.

Are you using any Asus utilities?

What are you using to cool the 5850? Is it a full cover block, and is it the correct one for your card? Have you checked ALL the temps on the card using GPU-Z?
June 17, 2011 4:36:51 AM

Proximon said:
How did you monitor voltages? Onboard sensors are useless. unless you have a good DMM you won't know. But of course a Seasonic PSU is not too suspect.

Are you using any Asus utilities?

What are you using to cool the 5850? Is it a full cover block, and is it the correct one for your card? Have you checked ALL the temps on the card using GPU-Z?



Voltages were originally monitored using the Asus Util (pre-OS wipe, they haven't been reinstalled). Wasn't aware that it was useless though. So I went back in verified the 12V and 5V rails with my DMM. Not sure how to get at the 3V while the machine is up and stress testing without making a jumper or something.

The 5850 is cooled with a full cover block. Rechecked at Sidewinder. Correct block. Stable temperatures on Burn-In run of FurMark off GPU-Z are 46C for the GPU and 62C for the VDDC's. And to be honest, I'm not seeing temps being the issue if AC:B goes down in seconds and Furmark will run for hours.

Appreciate the reply. Thanks.
Related resources
a c 113 B Homebuilt system
June 17, 2011 4:58:57 AM

Unless some of that video memory was only being accessed in games. That's what I'm wondering, if the block wasn't making contact with part of the memory on the GPU.

Vendors are very quick to release non-reference video cards, that often are not fully compatible with reference water blocks. I'm not saying that's your problem, just making sure you understand why I asked.

The next thing to suspect would be the drive, in a general sort of way. I would check for chipset driver errors by switching to the other ports... I think you have 6 Intel SATA ports and two Marvell?

Perhaps swap out the cable while you are at it.
June 17, 2011 5:38:36 AM

Proximon said:
Unless some of that video memory was only being accessed in games. That's what I'm wondering, if the block wasn't making contact with part of the memory on the GPU.

Vendors are very quick to release non-reference video cards, that often are not fully compatible with reference water blocks. I'm not saying that's your problem, just making sure you understand why I asked.

The next thing to suspect would be the drive, in a general sort of way. I would check for chipset driver errors by switching to the other ports... I think you have 6 Intel SATA ports and two Marvell?

Perhaps swap out the cable while you are at it.


No, I get what you're thinking of. At the more extreme end, I'm wondering whether to replace the video card first, or to replace the CPU/MB/RAM and then see if the video card needs to go too. I'm hesitant to give up the WC solution. I really, really like how quiet this rig is.

Anyway, I had the drive on an Intel port, swapped over to one of the Marvell's with a new cable. Same results.

Best solution

a c 113 B Homebuilt system
June 17, 2011 9:22:52 AM
Share

All kinds of other things to try first. Update the BIOS for instance.

There was some problem with the original P6T that caused them to release a V2, but I can't recall what it was.

Bumping the voltage slightly, on both the CPU and RAM, can't hurt.
June 17, 2011 3:59:37 PM

Proximon said:
All kinds of other things to try first. Update the BIOS for instance.

There was some problem with the original P6T that caused them to release a V2, but I can't recall what it was.

Bumping the voltage slightly, on both the CPU and RAM, can't hurt.


Should have been more precise. The MB is the V2. BIOS is the most current version.

I'll play with the voltages when I get home.
June 20, 2011 7:22:21 AM

Proximon said:
All kinds of other things to try first. Update the BIOS for instance.

There was some problem with the original P6T that caused them to release a V2, but I can't recall what it was.

Bumping the voltage slightly, on both the CPU and RAM, can't hurt.



And more, precisely appears to have solved the issue. Hopefully this will let the system run for at least another year before I actually _want_ to rebuild.

Thanks.
June 20, 2011 7:22:32 AM

Best answer selected by cdrikari.
a c 113 B Homebuilt system
June 21, 2011 1:49:11 AM

Glad we could help.
!